Dagu MCP Server

Content

Dagu is a local-first workflow engine for defining and operating YAML DAGs with a Web UI, single-binary deployment, and no required external DBMS. Its built-in MCP server exposes a compact Streamable HTTP interface for AI agents to read Dagu state, inspect DAG specs and logs, preview or apply workflow changes, and control DAG runs.

The MCP implementation registers three core tools: dagu_read for read-only state, specs, run details, and logs; dagu_change for previewing or applying DAG YAML changes; and dagu_execute for starting, enqueueing, retrying, or stopping DAG runs.

Source Review

• https://github.com/dagucloud/dagu
• https://github.com/dagucloud/dagu/blob/main/README.md
• https://github.com/dagucloud/dagu/blob/main/LICENSE
• https://github.com/dagucloud/dagu/blob/main/LICENSING.md
• https://github.com/dagucloud/dagu/blob/main/internal/service/mcp/server.go
• https://docs.dagu.sh/mcp/
• https://docs.dagu.sh/mcp/clients
• https://docs.dagu.sh/mcp/tools
• https://docs.dagu.sh/mcp/authentication
• https://docs.dagu.sh/mcp/auditability
• https://docs.dagu.sh/getting-started/installation
• https://docs.dagu.sh/overview/deployment-models
• https://docs.dagu.sh/web-ui/secrets
• https://github.com/dagucloud/dagu/releases
• https://github.com/dagucloud/dagu/pkgs/container/dagu

These sources were reviewed on 2026-06-06. Prefer the live repository, README, license notes, MCP server implementation, MCP docs, authentication docs, auditability docs, installation docs, deployment model docs, secrets docs, GitHub Releases, and GHCR package page for current setup, endpoint, tools, API key, and deployment behavior.

Features

• Expose a built-in Streamable HTTP MCP endpoint from the running Dagu server.
• Read DAG lists, DAG details, DAG specs, run details, run logs, and Dagu MCP reference resources.
• Preview or apply DAG YAML changes through dagu_change.
• Start, enqueue, retry, or stop workflow runs through dagu_execute.
• Use resources such as current DAG specs and run logs.
• Use prompts for creating DAGs, editing DAGs, and debugging failed runs.
• Scope API keys to the MCP surface when Dagu authentication is enabled.
• Keep MCP-triggered actions visible through Dagu's audit and Web UI surfaces.

Installation

Install Dagu through an upstream-supported path such as Homebrew, GitHub Releases, npm, Docker/GHCR, or Helm. For Homebrew:

brew install dagu

Start the Dagu server, enable authentication if required for your environment, and create an API key that is allowed to use the MCP surface. A sanitized MCP client configuration looks like:

{
  "mcpServers": {
    "dagu": {
      "url": "LOCAL_DAGU_MCP_URL",
      "headers": {
        "Authorization": "Bearer DAGU_MCP_API_KEY"
      }
    }
  }
}

Follow the upstream MCP clients and authentication docs for the exact local or remote endpoint, API key setup, and client-specific configuration.

Use Cases

• Let an AI coding agent inspect workflow state and logs before editing a DAG.
• Draft new Dagu YAML workflows from a natural-language goal, then preview validation before applying.
• Debug failed workflow runs by reading run metadata and logs through MCP.
• Start or enqueue approved workflow runs from Claude, Codex, Gemini CLI, or another MCP-capable agent.
• Operate scheduled automation from a local, self-hosted, managed, or hybrid Dagu deployment.

Safety and Privacy

Dagu can run real automation, including shell commands, Docker steps, Kubernetes Jobs, SSH actions, SQL queries, HTTP requests, file operations, notifications, and AI-agent harness steps. Treat dagu_change and dagu_execute as high-impact tools. Require review before applying workflow changes, launching production jobs, retrying expensive work, stopping incident response workflows, or changing schedules and credentials.

MCP clients can see DAG specs, logs, parameters, outputs, resource references, and audit context. Review log redaction, secrets management, API key scope, and model-provider exposure before giving an agent access to sensitive workflows or customer data.

Duplicate Check

No dagucloud/dagu entry, Dagu MCP Server entry, Dagu workflow MCP entry, or matching source URL was found in content/mcp.